diff --git a/docs/index.html b/docs/index.html
index 890ab58..b16136f 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -154,6 +154,7 @@ <h2>skval vs. <code>skill-creator</code></h2>
       <tr><td>Structural / lint checks</td><td><span class="dash">—</span></td><td><span class="A">✓</span> deterministic static checks <span class="dash">(D1)</span></td></tr>
       <tr><td>Safety gate</td><td><span class="dash">—</span></td><td><span class="A">✓</span> veto gate + safe extraction of untrusted skills <span class="dash">(D6)</span></td></tr>
       <tr><td>Offline, no-model path</td><td><span class="dash">— needs model runs + review</span></td><td><span class="A">✓</span> deterministic scan, CI-gateable (exit ≠ 0 on Reject)</td></tr>
+      <tr><td>Pre-run cost estimate</td><td><span class="dash">— tokens / time only, after the run</span></td><td><span class="A">✓</span> token + $ projection <em>before</em> you launch</td></tr>
       <tr><td>With-skill vs. baseline evals</td><td><span class="A">✓</span> core of the loop</td><td><span class="A">✓</span> effectiveness + lift, significance <span class="dash">(D2)</span></td></tr>
       <tr><td>Reliability</td><td>variance (mean ± stddev)</td><td><span class="A">✓</span> <code>pass^k</code> over N trials, scored <span class="dash">(D3)</span></td></tr>
       <tr><td>Triggering</td><td><span class="A">✓</span> description optimizer (train / test)</td><td><span class="A">✓</span> precision / recall / F1, scored <span class="dash">(D5)</span></td></tr>